IAN DEW-BECKER Northwestern University ROBERT J. GORDON Northwestern University Where Did the Productivity Growth Go? Inflation Dynamics and the Distribution of Income It’s no secret that the gap between the rich and the poor has been growing, but the extent to which the richest are leaving everybody else behind is not widely known.... It’s like chasing a speedboat with a rowboat. —Bob Herbert, The New York Times1 There is no question that a huge gap has opened up between productivity and living standards.... Not since World War II have productivity and income diverged so sharply. —Louis Uchitelle, The New York Times2 THE FIRST HALF of this decade has witnessed a sharp contrast between strong output and productivity growth, on the one hand, and slow employ- ment and median income growth, on the other. The strong growth in output combined with weak growth in hours worked has resulted in an explosion We are above all grateful to Daniel Feenberg of the National Bureau of Economic Research for making it possible for us to access and analyze the Internal Revenue Service micro data files. We are grateful for helpful data, ideas, and insights to Michael Harper and Phyllis Otto of the Bureau of Labor Statistics, Robert Brown of the Bureau of Economic Analysis, Carol Corrado of the Board of Governors of the Federal Reserve, Mary Daly, Bradford DeLong, Julie P. Gordon, Lawrence Katz, and Mark Watson. Daniel Feenberg and Emmanuel Saez were particularly helpful in providing insights and programs that made possi- ble our analysis of the IRS micro data files. Chris Taylor was our exemplary research assistant. 1. “The Mobility Myth,” The New York Times, June 6, 2005, p. A23. 2. “Were the Good Old Days That Good?” The New York Times, July 3, 2005, p. C1. 67 68 Brookings Papers on Economic Activity, 2:2005 in labor productivity growth, implying an underlying trend that is rising faster than in any previous subperiod of the postwar era. Yet who received the benefits? Median household income actually fell by 3.8 percent from 1999 to 2004 and grew from 1995 to 2004 at a rate of only 0.9 percent a year, a much slower rate than that of nonfarm private business (NFPB) output per hour over the same period, at 2.9 percent.3 Similarly, the median real wage for all workers grew over 1995–2003 at an average rate of 1.4 percent a year, less than half the rate of productivity growth.4 The failure of the productivity growth revival to boost the real incomes and wages of the median family and the median worker calls into question the standard economic paradigm that productivity growth translates auto- matically into rising living standards, as in this quote from Paul Krugman: Productivity isn’t everything, but in the long run it is almost everything. A coun- try’s ability to improve its standard of living over time depends almost entirely on its ability to raise its output per worker . the essential arithmetic says that long-term growth in living standards . depends almost entirely on productivity growth.5 This paper should be read in the spirit of a detective novel, whose title might be “The Case of the Missing Productivity Payoff.” Our detective story is divided into two parts, macro and micro. The macro part begins with the standard identity stating that productivity growth equals real wage growth if labor’s income share is constant. We examine the aggregate data that compare productivity growth with growth in alternative real wage measures, and we then ask how the post-1995 acceleration in productivity growth enters into the econometrics of price and wage dynamics. In past incarnations of dynamic Phillips curves, productivity growth has been a sideshow to the main story, if mentioned at all. The paper provides a new look at price and wage dynamics in order to assess the causes of low infla- tion in the decade after 1995. In light of the strong demand of the late 3. Mishel, Bernstein, and Allegretto (2005, table 1.1, p. 42). Median household income updated from 2003 to 2004 is from David Leonhardt, “U.S. Poverty Rate Was Up Last Year,” The New York Times, August 31, 2005, p. A1. 4. Mishel, Bernstein, and Allegretto (2005, table 2.6, p. 122). Measures of real income and real wages from this source deflate nominal values by the CPI-U-RS back to 1978. (The CPI-U-RS is a supplementary, research version of the CPI-U, which revises the CPI-U back to 1978 to incorporate improvements in measurement methodology since then.) 5. Krugman (1990, p. 9). Ian Dew-Becker and Robert J. Gordon 69 1990s, why was inflation so low? What role did the revival of trend produc- tivity growth play, in contrast to other beneficial supply shocks? Did the pro- ductivity growth slowdown of 1965–79 play a parallel role in creating high inflation in the 1970s? Can dynamic wage and price equations reproduce the behavior of the changes in labor’s income share observed in the data? The investigation begins with the Gordon inflation model that explains price changes by inertia, demand shocks, and supply shocks but does not include wages. A unique contribution of this paper is then to bring wages back into the study of inflation dynamics and to develop a model that includes both price and wage equations and allows each to feed back to the other. This model can capture the effect of changes in trend productivity growth on inflation, nominal wage changes, and changes in labor’s income share. In dynamic simulations of the wage-price model, we find that changes in the productivity growth trend had major effects in boosting inflation during 1965–79 and in slowing it down between 1995 and 2005. The second, micro part of the paper then examines the behavior of labor and nonlabor income as recorded in the micro data files of the Statistics of Income Division of the Internal Revenue Service (IRS) covering 1966–2001. The IRS micro files can be used to determine how much of the real income gain over various periods—for example, 1966–2001 or 1997–2001—accrued to taxpayers at the median income and to taxpayers at different percentiles of the income distribution from the 10th to the 99.99th. The IRS data have the great advantage over the more frequently used Current Population Sur- vey (CPS) data that they allow a microscopic view of what is going on inside the upper tenth of the income distribution. We find that increasing inequal- ity within the upper tenth is as important a source of growing inequality as the higher ratio of incomes in the top decile to those in the bottom decile. Our review of the sources of increased income inequality finds that economists have placed too much emphasis on skill-biased technical change and too little on independent factors that have pushed down rela- tive incomes at the bottom and raised them at the top. At the bottom we take a broader perspective that extends back to the 1920s, allowing us to explain the U-shaped time pattern of inequality by three U-shaped factors: the rise and fall of labor unions, the fall and rise of immigration, and the fall and rise of international trade. At the top we focus on the enormous increase in the income share of the top 1 percent and even the top 0.01 percent. Part of our analysis applies the “economics of superstars,” extending 70 Brookings Papers on Economic Activity, 2:2005 Sherwin Rosen’s original argument that new technology such as CDs, the Internet, and cable TV boosts the earnings premia of the very top perform- ers in certain fields.6 The other part examines the large share of added income in the upper tail of the income distribution attributable to CEOs and other top corporate officers, where the facts are clear but the interpre- tations are controversial. Taken together, the macro and micro parts of our detective story allow us to allocate across the income distribution the cumulative increase in real GDP attributable to the post-1995 acceleration in productivity growth. The macro analysis allows us to determine how much of the cumulative increase was broadly allocated to all income groups through lower infla- tion, how much went to nominal labor income, and how much went to nominal nonlabor income. The micro analysis allows us to look within the increase in real labor income to determine how much went to the top, mid- dle, and bottom of the income distribution. In the end we find that only the top 10 percent of taxpayers had gains in real labor income per hour that kept pace with productivity growth over either the 1966–2001 or the 1997–2001 period. The micro analysis thus reconciles the paradox that median income has lagged so far behind productivity growth while labor’s income share remained roughly constant, by showing that the distribu- tional change caused median income growth to behave very differently from mean income growth. Issues Raised by the Macro Data on Productivity and Labor’s Share of Income We begin by examining data on the interplay between productivity growth and labor’s share of income, exploiting the fact that, by definition, a constant income share of labor compensation implies that labor produc- tivity is growing at the same percentage rate as real labor compensation per hour. How large was the post-1995 acceleration in productivity growth? Did real wages respond by accelerating in equal measure, leaving labor’s income share intact, or did labor’s share decline? If so, how large was the difference across alternative measures of real wage growth? 6.